Altmetric

Euan Adie (euan@altmetric.com)
Jun 21st, 2012

Scientists talk. Let's listen.

We track and score scientific articles & datasets based on the attention they've received in social media, reference managers, news outlets and literature reviews.

Our partners license this data and present it to their end-users in compelling new ways and contexts:

• Supplementing article level metrics systems
• Adding to search indices
• Building new, user facing projects that have alt-metrics components

Altmetric Explorer

The "Explorer" web app is aimed at editors, press officers, science managers and funders interested in tracking the buzz surrounding relatively large sets of papers.

Ping me (Euan) and I'll give you an account if you'd like to take a look.

The Altmetric Explorer

• Browse attention paid to articles by journal or issue
• List influencers
• Search for specific articles
• See comparative ranking for articles by journal, DOI prefix, Medline subject area
• Automatically generate reports in Excel format
• Compare aggregated data for any two journals
• Set up email alerts for attention past a certain threshold

Priced at £50 p/m

Data access - jumpstart your alt-metrics applications.

Collecting alt-metrics data involves:

• Aggregating raw input
• Identifying the articles being discussed
• Disambiguating multiple identifiers (PMID, DOI, PII, handle, arXiv ID)
• Exposing the processed data through an API

... quickly and at scale.

Altmetric does all this for you - you can focus on adding context or presenting it to end users.

Aggregation is hard example - blogs

Simple blog aggregators are easy.

Building comprehensive (> 1k feeds) blog aggregation tools is non-trivial.

Services like Spinn3r are expensive ($30k p/a +).

Altmetric builds on experience gained developing postgenomic.com, nature.com blogs and scienceblogs.com.

Let us curate blog lists, fetch full text, weed out spam and handle disambiguation for you.

Using the data

The Altmetric dataset is available:

• As HTML embeds that can be placed directly on web pages, e.g.

  <script src='http://cdn.altmetric.com/embed.min.js'></script><div class='altmetric-embed' data-doi='10.1016/S0140-6736(11)61619-x'></div>

  

  Click on the score to see the conversation in a pop-up.

• Through an API (can be self-hosted)
• As bulk data downloads

Coverage

Coverage for articles with doi prefix 10.1186 published in December 2011 is ~ 44%

737 matched out of 1,655 total in PubMed (76k added that month)

261 (~ 1/3rd matched) had three or more posts (tweets, wall posts, G+ posts etc.)

Note that this includes any posts by BMC staff and bots.

What can you do with the raw data?

• Help users to discover new articles of interest
• ... and new people, journals and sites
• Measure impact in new, interesting ways
• Track the conversations around scholarly content
• Quantify the impact of non-traditional contributions and activities

See:

http://www.altmetric.com/interface/plos.html

http://www.applications.sciverse.com/action/appDetail/297955

The Altmetric API

The API accepts any identifier as input and returns:

• Any matching mentions
• Reference manager counts
• Historical context
• The Altmetric score

Try it out:

http://api.altmetric.com/unstable/doi/10.1242/jcs.033340

Data dump

If you're a bilbiometrics / altmetrics researcher we'd love to you take the Altmetric data and do with it as you see fit.

Here are some numbers and simple analyses to give you a feel for the dataset (and some idea of the size of the altmetrics universe as it applies to public shares).

Hit the right arrow to go to the next slide.

Lower-bound for number of scholarly articles shared online

Between June 14th and June 21st '12 Altmetric saw:

• 15,151 new papers (not seen previously)
• ~ 30,000 total papers mentioned

In context: avg. num new articles each fortnight in:

• PubMed ~ 39,000 articles
• RePEC ~ 6,000 articles (?)
• arXiv ~ 3,500 articles
• SSRN ~ 2,750 articles

So a significant fraction shared but prob < half.

How much sharing per article early on?

Look at 108k articles known to have been published so far this year

54% (59k) have only one mention

89% (96k) have 5 or fewer mentions

0.3% (427) have 100+ mentions

Where do scientists share public links to papers online?

Using data from the past two weeks...

How many people are sharing links to articles?

468,000 tweeters collected since Jul '11 (lower bound)

(in context: there are 6 - 10 million scientists worldwide)

62,358 (13%) shared more than one paper

6,933 (1.4%) shared more then ten papers

SNIP

"Created by Professor Henk Moed at CTWS, University of Leiden, Source-Normalized Impact per Paper (SNIP) measures contextual citation impact by weighting citations based on the total number of citations in a subject field. The impact of a single citation is given higher value in subject areas where citations are less likely, and vice versa." - http://www.journalmetrics.com/snip.php

What % of papers are shared?

I put this data together at 5am this morning. You should view it with skepticism.

Looked at all journals in PubMed Jan-Mar '12 with a SNIP in 2009, at least 10 papers with DOIs.